CN109003659A - Stomach Helicobacter pylori infects pathological diagnosis and supports system and method - Google Patents
Stomach Helicobacter pylori infects pathological diagnosis and supports system and method Download PDFInfo
- Publication number
- CN109003659A CN109003659A CN201710423850.6A CN201710423850A CN109003659A CN 109003659 A CN109003659 A CN 109003659A CN 201710423850 A CN201710423850 A CN 201710423850A CN 109003659 A CN109003659 A CN 109003659A
- Authority
- CN
- China
- Prior art keywords
- image data
- convolutional neural
- neural networks
- unit
- pathological
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Abstract
The invention discloses a kind of, and system and method are supported in the stomach Helicobacter pylori infection pathological diagnosis based on big data deep learning, the system includes: image data obtaining unit, and the pathological section image of the stomach Helicobacter pylori cases of infection for obtaining stomach normal slice image and having made a definite diagnosis is as input image data；Image data marks unit, for being labeled to input image data；Image data base construction unit, the classification of the image data of mark, arrangement for providing image data mark unit, constructs pathological image database；Convolutional neural networks (CNN) structural unit, for constructing the first convolution neural network model；And convolutional neural networks model training unit, obtain ideal convolutional neural networks model.Stomach Helicobacter pylori infection pathological diagnosis through the invention supports system and method that can realize accurate and efficient intelligent read tablet, the pathological diagnosis work infected with stomach Helicobacter pylori on adjuvant clinical, improves its accuracy rate, working efficiency and operation duration state.
Description
Technical field
The present invention relates to a kind of, and system is supported in the stomach Helicobacter pylori infection pathological diagnosis based on big data deep learning
And method.
Background technique
Deep learning most the agreeing with, be most widely used for image recognition and speech analysis that be current artificial intelligence field
Algorithm, inspiration from the working mechanism of human brain, be by establish convolutional neural networks to externally input data into
Row automation feature extraction obtains information and exports so as to make machine rational learning data.Currently, being based on deep learning
Artificial intelligence be applied to various industries field, including speech recognition, recognition of face, vehicle-logo recognition, handwritten Kanji recognition etc..
The research and development of products of artificial intelligence medical assistance technology also makes substantial progress in recent years, is such as ground by Google's brain and Verily company
The artificial intelligence product for breast cancer pathological diagnosis of hair can reach 89% tumor-localizing accuracy rate；Zhejiang University attached
One hospital realizes in quickly analysis thyroid gland B ultrasound the position of knuckle areas and good pernicious using artificial intelligence.
During medical diagnosis, histopathologic slide, which checks, needs high standardization and accuracy.Exhausted big portion at present
The histopathologic slide divided is to analyze by manual manufacture, and by pathologist in conjunction with the clinical diagnosis experience of itself long-term accumulation
And judgement.Modern medicine proves, helicobacter pylori (Helicobacter pylori, HP) is that chronic gastritis, digestibility are burst
The important virulence factor of ulcer, gastric cancer is classified as one of I class carcinogenic substance by international cancer research institution (IARC).Therefore, early stage is examined
Surveying HP infection, HP, treatment HP are diseases related to have important clinical meaning for eradicating in time.Clinically, it is taken by gastroscope pincers
Gastric mucosa tissue, specimens paraffin embedding slices, dyeing carry out histology microscopy and detect HP, are diagnosis for experienced Pathology Doctors '
" goldstandard " of bacterium infection.Compared to other HP detection methods, HP is directly detected using histopathological methods with following
Advantage: the 1. substantially lesion when gastroscope is drawn materials in clear stomach, such as ulcer, gastric cancer；2. determining stomach while clear HP infection
The degree and type of interior inflammation；3. lapsing to for Gastroduodenal disorders can be specified to the patient for receiving to check after HP eradication therapy；
4. clinical application takes 1 piece of sufficiently large sample that can diagnose 98% HP infection in antrum, positive evidence existing for HP is provided.Disease
The Warthern-Starry argentation of improvement can be used in reason slice, on 40X under the microscope visible faint yellow background in brown or
Brown bending or corynebacterium object, it is 5 μm long, in conjunction with HP multidigit in the feature in foveolae gastricae or intrinsic gland, it is easily recognized, positive rate
It is higher.Although as the development of modern medical techniques, rapid urease test method (RUT) and Noninvasive detection method such as blood
The clear detection methods such as detection, urea breath test of learning are more widely applied, but the above method can only reflect the feelings of HP infection indirectly
Condition, and respectively have limitation, for example RUT testing result is easy by patient medication history (such as recent application antibiotic, bismuth agent or proton pump
Inhibitor can lead to false negative result), the multifactor impacts such as position, experiment condition of drawing materials under gastroscope, and to have 10 in sample4With
On bacterium just show the positive.
Therefore histopathological methods detection HP can intuitively look into the sensitivity for seeing pathogen and special detection hand as a kind of
Section still has highly important clinical value for HP infection and diseases related diagnosis, treatment and follow-up.Therefore, pathology
The shortcomings that histology artificial detection HP infects is mainly: the result judgement of pathology slide is that gained is visually observed by pathologist,
The subjective factors such as this artificial diagosis method and pathologist experience, working condition are closely related, are easy to produce error.To the greatest extent
The form that pipe improves helicobacter pylori under W-S is dyed is clear but few for bacterial number, chipping qualities is irregular, stomach lining
Possibly even under the complex situations such as interference pathogen observation, the insufficient pathologist of experience is easy to fail to pinpoint a disease in diagnosis, miss the complicated multiplicity of lesion
It examines.Meanwhile pathologist will be responsible for checking all visible biological tissues on slice, and each patient can there are many cut
Piece, when carrying out 40 times of amplifications, each slice has more than 100 hundred million pixel, therefore the workload of artificial diagosis is very big, be easy by
The influence of the factors such as diagosis person's subjective emotion and tired diagosis.Moreover, different virologists may provide phase to same patient
When different diagnosis.Therefore, the Tissue pathological diagnosis method that this height relies on human factor has subjective differences,
In addition the disadvantages of its great work intensity, time cost height and diagnosis inconsistency, largely will affect the early stage of HP infection
Diagnosis and treatment are to influence patient's prognosis.In addition, the pathologist of the qualified profession of culture needs to carry out long-term professional training and practice
Process, cultivation cycle is long, and the influence vulnerable to social factors such as current social economy, culture, it is meant that China or even the whole world
The big severe situation urgent need to resolve of pathologist quantity " supply falls short of demand ", professional notch.
Summary of the invention
The shortcomings that diagosis artificial for Histopathology, the quasi- computer that passes through of the present invention is to a large amount of stomach Helicobacter pylori senses
It contaminates pathological image and carries out deep learning, to establish intelligentized stomach Helicobacter pylori infection pathological diagnosis mathematical model, build
Auxiliary pathological diagnosis artificial intelligence platform is infected based on the stomach Helicobacter pylori of big data and deep learning algorithm, to realize
High-accuracy and efficient intelligent read tablet, the pathological diagnosis work infected with stomach Helicobacter pylori on adjuvant clinical improve
Its accuracy rate, working efficiency and operation duration state.
Based on this, it is an object of the invention to overcome above-mentioned the deficiencies in the prior art place one kind is provided, clinic can be improved
System is supported in the stomach Helicobacter pylori infection pathological diagnosis of efficiency, reduction medical treatment cost when Diagnosis of Gastric Helicobacter pylori infection
System.
To achieve the above object, the technical scheme adopted by the invention is as follows: a kind of stomach pylorus based on big data deep learning
System is supported in helicobacter infection pathological diagnosis, and the support system includes: image data obtaining unit, normal for obtaining stomach
Sectioning image and the pathological section image for the stomach Helicobacter pylori cases of infection made a definite diagnosis are as input image data；Image
Data mark unit, for being labeled to the input image data, and guarantee image label and image it is true
Pathological diagnosis result is consistent；Image data base construction unit, the mark figure for providing described image data mark unit
It as data classification, arranges, constructs pathological image database；Convolutional neural networks structural unit, for constructing the first convolutional Neural
Network model；And convolutional neural networks model training unit, using the image data of the pathological image database to described
The parameter of first convolution neural network model is adjusted, and training the first convolution neural network model, can be used
In the second convolution neural network model of detection patient's pathology image data.
Doctor is in combination with the holding equipment for the classification results that provide of patient's pathological image of input and corresponding as a result,
Probability and doctor professional standing and experience be rapidly diagnosed to be the patient whether with stomach Helicobacter pylori sense
Dye, significantly improves the efficiency of clinical diagnosis, to reduce medical treatment cost；Wherein, in order to guarantee that the image data being collected into is accurate
It is errorless, it can use image labeling tool ASAP, every pathological section image be labeled, to guarantee the label of image and true
Real value is consistent；In order to accelerate train network model speed, can be used with high-speed parallel calculate GPU come replace CPU into
Row training；It, can be by training based on convolutional neural networks training unit in order to accelerate the detection speed of convolutional neural networks model
Good network model is modeled as the CNN disaggregated model structure of variable step size again, for the detection method in practical operation；It should
Model will carry out blocking processing to huge full slice image, and the living tissue region segmentation selected in advance is identical at size
ROI piecemeal, due to the detection between piecemeal can with highly-parallel so that the speed of detection more GPU it is parallel under obtain it is significant
It is promoted, then by the detection of the CNN disaggregated model of variable step size, generates prediction probability distributed image；Image data base will be sick
Reason image data is divided into training set, test set and checksum set etc.；The parameter of first convolution neural network model include learning rate,
The network parameters such as frequency of training and how many layer network, trained to refer to when seeking optimal solution, the process of automatically adjusting parameter.
Preferably, the support system further includes convolutional neural networks model testing unit, for obtaining ideal convolution
Neural network model.It should be noted that " ideal " refers to that the accuracy rate of convolutional neural networks model is high herein, and " Shandong
Stick ".
Preferably, the convolutional neural networks model testing unit includes convolutional neural networks model checking unit and convolution
Neural network model test cell, the convolutional neural networks model checking unit is for detecting second convolutional neural networks
The accuracy rate of model；The convolutional neural networks model measurement unit is for detecting the second convolution neural network model
No over-fitting, to filter out the third convolutional neural networks model of robust；If should be noted that model on test set
Accuracy rate differs larger with the accuracy rate in checksum set training, then illustrates model over-fitting, at this point, can return to convolutional neural networks
In training unit, regulating networks structure or parameter train again to obtain better network model；If on test set
Accuracy rate and checksum set training in accuracy rate it is very close, then illustrate the model more robust.
Preferably, the support system further includes convolutional neural networks model pre-training unit, for working as described image number
When the deficiency of input image data being collected into according to obtaining unit, pre-training is carried out to the first convolution neural network model.
Preferably, the support system further includes pathological image data pre-processing unit, for screening and showing disease
Manage the area to be tested in image.
Preferably, in order to ensure the validity of detection, the pretreatment unit is filtered out described using Adaptive Thresholding
Area to be tested.
Preferably, the convolutional neural networks training unit is using fine tuning method training the first convolutional neural networks mould
Type.
As another aspect of the present invention, the present invention also provides a kind of pathological diagnosis branch of stomach Helicobacter pylori infection
Hold method, the support method the following steps are included:
Image data obtains: obtaining the pathology of stomach normal slice image and the stomach Helicobacter pylori cases of infection made a definite diagnosis
Sectioning image is as input image data；
Image data mark: being labeled the input image data, and guarantees the label and image of image
True pathological diagnosis result is consistent；
Image data base building: the classification of the image data of mark, the arrangement that described image data mark unit is provided, structure
Build pathological image database；
Convolutional neural networks construction: the first convolution neural network model of construction；And
Convolutional neural networks model training: using the image data of the pathological image database to the first convolution mind
Parameter through network model is adjusted, and training the first convolution neural network model, acquisition can be used for detecting patient
Second convolution neural network model of pathological image data.
It should be noted that image data mark and image data base building are considered as pathological image database sharing rank
Section.Preferably, the support method further includes convolutional neural networks model testing step: obtaining ideal convolutional neural networks mould
Type；The convolutional neural networks model testing step includes that convolutional neural networks model checking and convolutional neural networks model are surveyed
Examination, the convolutional neural networks model checking are used to detect the accuracy rate of the second convolution neural network model；The convolution
Neural network model test, for detect the second convolution neural network model whether over-fitting, to filter out the of robust
Three convolutional neural networks models.It should be noted that convolutional neural networks construction, convolutional neural networks model training and convolution mind
The training stage that can regard convolutional neural networks as is examined through network model, for obtaining ideal convolutional neural networks model.
As the third aspect of the invention, the invention further relates to above-mentioned support systems in pathological diagnosis stomach H. pylori
Clinical application in bacterium infection.
In conclusion the invention has the benefit that
Compared with the artificial diagosis of existing pathologist, the present invention is based on the stomach pyloruses of big data and deep learning algorithm
The holding equipment of helicobacter infection pathological diagnosis has the advantages that accuracy rate is high, time-consuming short and run duration is long, and this
Invention includes that front three, the popularization of basic hospital and cloud service will be helpful to solve medical resource distribution unevenness in various big hospital
Even, long-range high-quality medical treatment of realization etc., provides more convenient, more accurate pathological diagnosis service for many patients；The reality of above-mentioned advantage
It is now to allow computer to carry out big data because device and method of the invention utilize deep learning algorithm in the advantage of image recognition
The deep learning of the stomach Helicobacter pylori infection pathological section of rank, so that pathologist diagosis and therewith can be simulated by training
The intelligent neural network model to match in excellence or beauty, by constantly learning and verifying, which be may be implemented to stomach pylorus spiral shell
The intelligent diagosis of bacillus infection pathological section is revolved, quickly identifies and obtains scientific conclusion.
Detailed description of the invention
Fig. 1 is the structural block diagram that stomach Helicobacter pylori of the invention infects that system is supported in pathological diagnosis；
Fig. 2 is that the stomach Helicobacter pylori of the embodiment of the present invention two infects the flow chart that method is supported in pathological diagnosis；
Fig. 3 is to infect slice map to stomach Helicobacter pylori；
Fig. 4 is the schematic diagram of the quick detection model of the embodiment of the present invention two；
Fig. 5 is that stomach Helicobacter pylori of the invention infects flow chart of the pathological diagnosis support system in application；
Wherein, 1, stomach Helicobacter pylori infection pathological diagnosis support system, 2, image data obtaining unit, 3, picture number
According to mark unit, 4, convolutional neural networks structural unit, 5, convolutional neural networks model training unit, 6, convolutional neural networks mould
Type verification unit, 7, image data base construction unit, 8, pathological image data pre-processing unit, 9, input terminal, 10, output end
End.
Specific embodiment
To better illustrate the object, technical solutions and advantages of the present invention, below in conjunction with the drawings and specific embodiments pair
The present invention is described further.
Embodiment 1
Referring to Fig. 1, a kind of embodiment of system 1, packet are supported in stomach Helicobacter pylori infection pathological diagnosis of the invention
It includes:
Image data obtaining unit 2, for obtaining stomach normal slice image and the stomach Helicobacter pylori made a definite diagnosis infection
The pathological section image of case is as input image data；
Image data marks unit 3, for being labeled to input image data, and guarantees the label and figure of image
The true pathological diagnosis result of picture is consistent；
Image data base construction unit 7, the image data of mark for providing image data mark unit classify, are whole
Reason constructs pathological image database；
Convolutional neural networks structural unit 4, for constructing the first convolution neural network model；
Convolutional neural networks model training unit 5, using the image data of pathological image database to the first convolutional Neural
The parameter of network model is adjusted, and the first convolution neural network model of training, acquisition can be used for detecting patient's pathology figure
As the second convolution neural network model of data；
Convolutional neural networks model testing unit 6, for obtaining ideal convolutional neural networks model, including convolutional Neural
Network model verification unit (not shown) and convolutional neural networks model measurement unit (not shown), convolutional Neural net
Network model checking unit is used to detect the accuracy rate of the second convolution neural network model；Convolutional neural networks model measurement unit,
For detect the second convolution neural network model whether over-fitting, to filter out the third convolutional neural networks model of robust.
Convolutional neural networks model pre-training unit (not shown), for what is be collected into when image data obtaining unit
When input image data deficiency, pre-training is carried out to the first convolution neural network model；And
Pathological image data pre-processing unit 8, for screening and showing the area to be tested in patient's pathological image.
Wherein, pathological image data pre-processing unit 8 filters out area to be tested using Adaptive Thresholding；Convolutional Neural
Network model training unit 5 is using fine tuning (fine-tune) method the first convolution neural network model of training；Include in database with
Lower four class data sets: training set, disappear test collection, test set and routine pathological image data set is disclosed.
In addition, input terminal 9 is used to infect existing stomach normal slice image and the stomach Helicobacter pylori made a definite diagnosis
The pathological section image input image data obtaining unit 2 of case, also, the data of these inputs finally will be by image data base
7 categorised collection of construction unit, for supporting subsequent clinical diagnosis to work；
And the pathological section image of patient to be detected is inputted into pathological image data pre-processing unit 8；
Outlet terminal 10, the convolutional neural networks of the robust for will be obtained by convolutional neural networks model training unit 5
Result of the model to the pathological section image classification detection of the patient to be detected of input pathological image data pre-processing unit 8
(histological type and corresponding probability) is presented to doctor, for clinical diagnosis reference.
Embodiment 2
Referring to fig. 2, stomach Helicobacter pylori Infect And Diagnose of the invention supports a kind of embodiment of method comprising as follows
Step:
(1) image data is acquired
Using ZhongShan University attached No.6 Hospital medical biotechnology library data as data source, 14000 pathological section figures are acquired
Picture, including 7000 stomach normal tissue sections images and 7000 stomach Helicobacter pylori infected tissues slice, and respectively according to
Training set: checksum set: test set=3:1:1 quantitative proportion is grouped at random.It is as shown in table 1 below:
The specific data of 1 pathological section image of table.
Acquired image is subjected to digital scanning storage, serial number is filed, creation stomach Helicobacter pylori infection pathology
Image data base.
(2) image information is marked
Using existing ASAP image labeling software to the disease of training set collected by step (1), checksum set and test set
It manages sectioning image and carries out data markers.For the accuracy for guaranteeing information labeling, processing need to be optimized to image before mark.It is right
The mark work of image, which is specifically included that, sketches the contours of various pathologic structure regions in image with different colours/thickness/actual situation lines,
According to the substantially lesion in stomach, such as ulcer, gastric cancer；While HP infects, the degree and type of inflammation in stomach are determined；Under the microscope
It can be seen that in brown or brown bending or corynebacterium object in faint yellow background；HP multidigit in the situations such as in foveolae gastricae or intrinsic gland,
Then to image classification and score value is assigned, and the region sketched the contours is subjected to label name.By the pathological image after correct mark
Digitlization storage is carried out, to carry out the network model training and verification of next step.Fig. 3 be to stomach Helicobacter pylori infection in not
The mark figure in typical hyperplasia region.
(3) training convolutional neural networks
1. designing a model
(a) convolutional Neural is constructed according to convolutional layer, maximum sample level, nonlinear function, the cascade mode of full articulamentum
Network；
(b) enhance the capability of fitting of network using multitiered network；
(c) port number of the output of the last full articulamentum of network is set as 2, and respectively representing the image is stomach normal slice figure
Picture, stomach Helicobacter pylori infected tissue sectioning image.
2. training network
(a) according to the image data being collected into step (1), (2), the parameter of convolutional neural networks model is adjusted
Section, observes the accuracy rate of classification on checksum set；
(b) in order to accelerate train network speed, replace CPU to be trained using the GPU calculated with high-speed parallel；
(c) method of the update of convolutional neural networks weighting parameter is solved using SGD, if convergence rate is slower, is made
With Adadelta, the optimization methods such as Adam are solved；
If training data (i.e. image data) number that (d) step (1) is collected into is very little, convolutional neural networks model is adopted
Fine-tune (fine tuning) is used in conventional open pathological image data set pre-training, then by the image data being collected into elder generation
Method carry out training convolutional neural networks model；
(e) training, the accuracy rate of classification can not rise such as on existing convolutional neural networks model, can be rolled up by increasing
The depth of neural network network model is accumulated to increase the capability of fitting of convolutional neural networks model.
3. designing quick detection model (as shown in Figure 4)
1. using Adaptive Thresholding in order to improve detection efficiency in pretreatment stage, being selected in advance from full slice image
Living tissue region, the test object (as shown in Fig. 4 arrow 101, representing preprocessing process) as convolutional neural networks.
2. in order to improve the accuracy of detection and flexibility can be by trained convolutional Neural net based on step (3)
Network is modeled as the CNN disaggregated model of variable step size again, for the detection method in practical operation；The model will be to huge
Full slice image carries out blocking processing, by the living tissue region segmentation selected in advance at the identical ROI piecemeal of size；Due to dividing
Detection between block can be with highly-parallel, so that the speed of detection is effectively promoted (such as Fig. 4 arrow in the case where more GPU
Shown in first 102, the quick detection process of representative model).By the detection of convolutional neural networks model, prediction probability distribution map is generated
Picture.
3. the prediction probability distributed image based on step 2, in post-processing, after screening out scattered point, analysis prediction probability point
Butut, to obtain the prediction result (as shown in Fig. 4 arrow 103, representing last handling process) of final full slice image.
(4) test set is verified
(a) the disaggregated model structure of the variable step size based on step 3. uses trained convolutional Neural net in step (3)
Network model tests test set, accuracy rate of the observing and nursing on test set.
If (b) in step (3) trained convolutional neural networks model test set upper accuracy rate and training in
The accuracy rate difference of checksum set is larger, then illustrates model over-fitting；At this point, can return in step (3), convolutional neural networks are adjusted
Prototype network structure or parameter obtain better network model.
If (c) in step (3) in accuracy rate of the trained convolutional neural networks model on test set and training
The accuracy rate of checksum set is very close, then illustrates the resulting convolutional neural networks model of the training more robust, and it is suitable to can be used as
Detection sufferer pathological image network model.
Embodiment 3
A kind of application examples of system is supported in stomach Helicobacter pylori infection pathological diagnosis of the invention, by pathology to be detected
Image is pre- by the pathological image data that input terminal 9 inputs in stomach Helicobacter pylori Infect And Diagnose holding equipment of the invention
Processing unit 8, operational process later is referring to Fig. 5, wherein
(a) in order to ensure the validity of detection, Adaptive Thresholding is used in the incipient stage, is preselected from full slice image
Living tissue region out is then based on threshold value results area Main subrack and selects area to be tested (i.e. patient pathologic tissue areas)；
(b) after, patient pathologic tissue areas picture is pre-processed, pretreatment includes denoising, histogram equalization, returns
One change and etc.；
(c) right with previous trained convolutional neural networks model (i.e. the second convolution neural network model in embodiment 1)
Area to be tested carries out classification and Detection in pretreated picture, thus obtain the prediction result of stomach Helicobacter pylori infection,
Classification and corresponding probability are infected including stomach Helicobacter pylori belonging to the pathological section.
The stomach Helicobacter pylori infection pathological diagnosis of the invention of embodiment 4 supports method compared with existing method
Clinically pathological diagnosis work is by cutting by pathologist manual read's pathological tissue of standardized training at present
Piece makes analysis and diagnosis in conjunction with the clinical diagnosis experience of itself long-term accumulation.Due to this artificial eye diagosis method with
The factors such as pathologist experience, working condition, subjective emotion are closely related, therefore accuracy rate is not high, but time-consuming, and work is held
Continuous limited time, be easy to produce fail to pinpoint a disease in diagnosis, situations such as mistaken diagnosis and diagnosis are inconsistent.The present invention then utilizes computer to standardized big
The deep learning for measuring stomach Helicobacter pylori infection pathological image carries out parameter regulation to convolutional neural networks and fitting is trained,
To obtain the network model of more robust.This neural network based on big data and deep learning can simulate artificial diagosis,
Corresponding output valve i.e. pathological diagnosis conclusion is obtained according to the new pathological image of input.Furthermore it by Model Reconstruction, is not influencing
In the case where accuracy in detection, detection speed is greatly improved.
30 doctors with 3 years or more stomach Helicobacter pylori Infect And Diagnoses and Couple herbs are chosen, everyone mentions respectively
For the pathological image of 40 doubtful stomach Helicobacter pylori infection, its type is judged, then calculate accuracy rate and average time, system
Diagnosis state is counted, compared with diagnosis support method of the invention, result is as shown in table 2 below.
The comparison of 2 stomach Helicobacter pylori Infect And Diagnose result of table
From upper table 2 it is found that reading histopathologic slide using method of the invention, accuracy rate is than professional pathologist
Higher level, and it is time-consuming shorter, run duration is long.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention rather than protects to the present invention
The limitation of range is protected, although the invention is described in detail with reference to the preferred embodiments, those skilled in the art should
Understand, it can be with modification or equivalent replacement of the technical solution of the present invention are made, without departing from the essence of technical solution of the present invention
And range.
Claims (10)
1. stomach Helicobacter pylori, which infects pathological diagnosis, supports system, which is characterized in that the support system includes:
Image data obtaining unit, stomach Helicobacter pylori cases of infection for obtaining stomach normal slice image and having made a definite diagnosis
Pathological section image is as input image data；
Image data marks unit, for being labeled to the input image data, and guarantees the label and figure of image
The true pathological diagnosis result of picture is consistent；
Image data base construction unit, the image data of mark for providing described image data mark unit classify, are whole
Reason constructs pathological image database；
Convolutional neural networks structural unit, for constructing the first convolution neural network model；And
Convolutional neural networks model training unit, using the image data of the pathological image database to the first convolution mind
Parameter through network model is adjusted, and training the first convolution neural network model, acquisition can be used for detecting patient
Second convolution neural network model of pathological image data.
2. support system according to claim 1, which is characterized in that the support system further includes convolutional neural networks mould
Type verification unit, for obtaining ideal convolutional neural networks model.
3. support system according to claim 2, which is characterized in that the convolutional neural networks model testing unit includes
Convolutional neural networks model checking unit and convolutional neural networks model measurement unit, the convolutional neural networks model checking list
Member is for detecting the accuracy rate of the second convolution neural network model；The convolutional neural networks model measurement unit, is used for
Detect the second convolution neural network model whether over-fitting, to filter out the third convolutional neural networks model of robust.
4. support system according to claim 1, which is characterized in that the support system further includes convolutional neural networks mould
Type pre-training unit, when the deficiency of input image data for being collected into when described image data acquiring unit, to described
One convolution neural network model carries out pre-training.
5. support system according to claim 1, which is characterized in that the support system further includes that pathological image data are pre-
Processing unit, for screening and showing the area to be tested in patient's pathological image.
6. support system according to claim 5, which is characterized in that the pretreatment unit is sieved using Adaptive Thresholding
Select the area to be tested.
7. support system according to claim 1, which is characterized in that the convolutional neural networks training unit is using fine tuning
Method training the first convolution neural network model.
8. a kind of support method of stomach Helicobacter pylori infection pathological diagnosis, which is characterized in that the support method include with
Lower step:
Image data obtains: obtaining the pathological section of stomach normal slice image and the stomach Helicobacter pylori cases of infection made a definite diagnosis
Image is as input image data；
Image data mark: the input image data is labeled, and guarantee image label and image it is true
Pathological diagnosis result is consistent；
Image data base building: the classification of the image data of mark, the arrangement that described image data mark unit is provided, building disease
Manage image data base；
Convolutional neural networks construction: the first convolution neural network model of construction；And
Convolutional neural networks model training: using the image data of the pathological image database to the first convolution nerve net
The parameter of network model is adjusted, and training the first convolution neural network model, acquisition can be used for detecting patient's pathology
Second convolution neural network model of image data.
9. support method according to claim 8, which is characterized in that the support method further includes convolutional neural networks mould
Type checking procedure: ideal convolutional neural networks model is obtained；The convolutional neural networks model testing step includes convolution mind
Through network model verification and convolutional neural networks model measurement, the convolutional neural networks model checking is for detecting described second
The accuracy rate of convolutional neural networks model；The convolutional neural networks model measurement, for detecting the second convolution nerve net
Network model whether over-fitting, to filter out the third convolutional neural networks model of robust.
10. application of -6 any holding equipments in Diagnosis of Gastric Helicobacter pylori infection according to claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710423850.6A CN109003659A (en) | 2017-06-07 | 2017-06-07 | Stomach Helicobacter pylori infects pathological diagnosis and supports system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710423850.6A CN109003659A (en) | 2017-06-07 | 2017-06-07 | Stomach Helicobacter pylori infects pathological diagnosis and supports system and method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109003659A true CN109003659A (en) | 2018-12-14 |
Family
ID=64572920
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710423850.6A Withdrawn CN109003659A (en) | 2017-06-07 | 2017-06-07 | Stomach Helicobacter pylori infects pathological diagnosis and supports system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109003659A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111210451A (en) * | 2019-11-29 | 2020-05-29 | 苏州优纳医疗器械有限公司 | Method for extracting helicobacter pylori form in all-digital slice image |
| CN112651375A (en) * | 2021-01-05 | 2021-04-13 | 中国人民解放军陆军特色医学中心 | Helicobacter pylori stomach image recognition and classification system based on deep learning model |
| CN114678121A (en) * | 2022-05-30 | 2022-06-28 | 上海芯超生物科技有限公司 | HP spherical deformation diagnosis model and construction method thereof |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103886318A (en) * | 2014-03-31 | 2014-06-25 | 武汉天仁影像科技有限公司 | Method for extracting and analyzing nidus areas in pneumoconiosis gross imaging |
| CN103971342A (en) * | 2014-05-21 | 2014-08-06 | 厦门美图之家科技有限公司 | Image noisy point detection method based on convolution neural network |
| CN104850825A (en) * | 2015-04-18 | 2015-08-19 | 中国计量学院 | Facial image face score calculating method based on convolutional neural network |
| US20150238148A1 (en) * | 2013-10-17 | 2015-08-27 | Siemens Aktiengesellschaft | Method and system for anatomical object detection using marginal space deep neural networks |
| CN104992177A (en) * | 2015-06-12 | 2015-10-21 | 安徽大学 | Internet porn image detection method based on deep convolution nerve network |
| CN105117611A (en) * | 2015-09-23 | 2015-12-02 | 北京科技大学 | Determining method and system for traditional Chinese medicine tongue diagnosis model based on convolution neural networks |
| CN105118068A (en) * | 2015-09-29 | 2015-12-02 | 常熟理工学院 | Medical image automatic annotation method under small sample condition |
| CN105160361A (en) * | 2015-09-30 | 2015-12-16 | 东软集团股份有限公司 | Image identification method and apparatus |
| CN105447866A (en) * | 2015-11-22 | 2016-03-30 | 南方医科大学 | X-ray chest radiograph bone marrow suppression processing method based on convolution neural network |
| CN105931224A (en) * | 2016-04-14 | 2016-09-07 | 浙江大学 | Pathology identification method for routine scan CT image of liver based on random forests |
| CN105975793A (en) * | 2016-05-23 | 2016-09-28 | 麦克奥迪（厦门）医疗诊断系统有限公司 | Auxiliary cancer diagnosis method based on digital pathological images |
| CN106529468A (en) * | 2016-11-07 | 2017-03-22 | 重庆工商大学 | Finger vein identification method and system based on convolutional neural network |
| CN106529116A (en) * | 2015-09-11 | 2017-03-22 | 西门子保健有限责任公司 | Diagnostic system and diagnostic method |
| CN106803247A (en) * | 2016-12-13 | 2017-06-06 | 上海交通大学 | A kind of microaneurysm automatic testing method based on multistage screening convolutional neural networks |
-
2017
- 2017-06-07 CN CN201710423850.6A patent/CN109003659A/en not_active Withdrawn
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150238148A1 (en) * | 2013-10-17 | 2015-08-27 | Siemens Aktiengesellschaft | Method and system for anatomical object detection using marginal space deep neural networks |
| CN103886318A (en) * | 2014-03-31 | 2014-06-25 | 武汉天仁影像科技有限公司 | Method for extracting and analyzing nidus areas in pneumoconiosis gross imaging |
| CN103971342A (en) * | 2014-05-21 | 2014-08-06 | 厦门美图之家科技有限公司 | Image noisy point detection method based on convolution neural network |
| CN104850825A (en) * | 2015-04-18 | 2015-08-19 | 中国计量学院 | Facial image face score calculating method based on convolutional neural network |
| CN104992177A (en) * | 2015-06-12 | 2015-10-21 | 安徽大学 | Internet porn image detection method based on deep convolution nerve network |
| CN106529116A (en) * | 2015-09-11 | 2017-03-22 | 西门子保健有限责任公司 | Diagnostic system and diagnostic method |
| CN105117611A (en) * | 2015-09-23 | 2015-12-02 | 北京科技大学 | Determining method and system for traditional Chinese medicine tongue diagnosis model based on convolution neural networks |
| CN105118068A (en) * | 2015-09-29 | 2015-12-02 | 常熟理工学院 | Medical image automatic annotation method under small sample condition |
| CN105160361A (en) * | 2015-09-30 | 2015-12-16 | 东软集团股份有限公司 | Image identification method and apparatus |
| CN105447866A (en) * | 2015-11-22 | 2016-03-30 | 南方医科大学 | X-ray chest radiograph bone marrow suppression processing method based on convolution neural network |
| CN105931224A (en) * | 2016-04-14 | 2016-09-07 | 浙江大学 | Pathology identification method for routine scan CT image of liver based on random forests |
| CN105975793A (en) * | 2016-05-23 | 2016-09-28 | 麦克奥迪（厦门）医疗诊断系统有限公司 | Auxiliary cancer diagnosis method based on digital pathological images |
| CN106529468A (en) * | 2016-11-07 | 2017-03-22 | 重庆工商大学 | Finger vein identification method and system based on convolutional neural network |
| CN106803247A (en) * | 2016-12-13 | 2017-06-06 | 上海交通大学 | A kind of microaneurysm automatic testing method based on multistage screening convolutional neural networks |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111210451A (en) * | 2019-11-29 | 2020-05-29 | 苏州优纳医疗器械有限公司 | Method for extracting helicobacter pylori form in all-digital slice image |
| CN112651375A (en) * | 2021-01-05 | 2021-04-13 | 中国人民解放军陆军特色医学中心 | Helicobacter pylori stomach image recognition and classification system based on deep learning model |
| CN114678121A (en) * | 2022-05-30 | 2022-06-28 | 上海芯超生物科技有限公司 | HP spherical deformation diagnosis model and construction method thereof |
| CN114678121B (en) * | 2022-05-30 | 2022-09-09 | 上海芯超生物科技有限公司 | Method and system for constructing HP spherical deformation diagnosis model |
| WO2023231176A1 (en) * | 2022-05-30 | 2023-12-07 | 上海芯超生物科技有限公司 | Construction method and system for hp spherical deformation diagnosis model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107368670A (en) | Stomach cancer pathology diagnostic support system and method based on big data deep learning | |
| CN107368671A (en) | System and method are supported in benign gastritis pathological diagnosis based on big data deep learning | |
| CN107369151A (en) | System and method are supported in GISTs pathological diagnosis based on big data deep learning | |
| Orlando et al. | An ensemble deep learning based approach for red lesion detection in fundus images | |
| CN106056595B (en) | Based on the pernicious assistant diagnosis system of depth convolutional neural networks automatic identification Benign Thyroid Nodules | |
| CN108717554A (en) | A kind of thyroid tumors histopathologic slide image classification method and its device | |
| CN109670510B (en) | Deep learning-based gastroscope biopsy pathological data screening system | |
| CN108694718A (en) | The same period new chemoradiation therapy curative effect evaluation system and method before rectal cancer | |
| Mendonça et al. | PH 2-A dermoscopic image database for research and benchmarking | |
| Hao et al. | Fusing multiple deep models for in vivo human brain hyperspectral image classification to identify glioblastoma tumor | |
| CN109858540B (en) | Medical image recognition system and method based on multi-mode fusion | |
| CN109903280A (en) | Tumour determines system, method and storage medium | |
| JP2023544466A (en) | Training method and device for diagnostic model of lung adenocarcinoma and squamous cell carcinoma based on PET/CT | |
| CN112085113B (en) | Severe tumor image recognition system and method | |
| CN107408198A (en) | The classification of cell image and video | |
| CN109255354A (en) | medical CT-oriented computer image processing method and device | |
| CN108352062A (en) | Method and apparatus for tissue identification | |
| CN109003659A (en) | Stomach Helicobacter pylori infects pathological diagnosis and supports system and method | |
| CN110946552B (en) | Cervical cancer pre-lesion screening method combining spectrum and image | |
| Maduri et al. | Malaria detection using image processing and machine learning | |
| CN113539476A (en) | Stomach endoscopic biopsy Raman image auxiliary diagnosis method and system based on artificial intelligence | |
| Swiderska-Chadaj et al. | Convolutional neural networks for lymphocyte detection in immunohistochemically stained whole-slide images | |
| CN114529554A (en) | Intelligent auxiliary interpretation method for gastric cancer HER2 digital pathological section | |
| Aubreville et al. | Field of interest prediction for computer-aided mitotic count | |
| CN105989245A (en) | Method and apparatus for obtaining values of characteristic quantification parameters of fibrosis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20181214 |
|
| WW01 | Invention patent application withdrawn after publication |